 Yongye ZhuandClaude Opus 4.8
|
fa248139a0
|
[MoE] Plumb gemm1_alpha/beta/clamp_limit into TRT-LLM FP8 MoE (#45723)
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
2026-07-01 14:34:05 -07:00 |
|
|
Yongye Zhu
|
d3229431f9
|
[DSV4] Better MXFP8 quantization kernel (#47229)
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
|
2026-07-01 14:33:51 -07:00 |
|
|
Harry Mellor
|
e7d0fcbc09
|
[CI] Fix various failures on main (#47197)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
|
2026-07-01 10:35:34 +01:00 |
|
|
|
ed41aa270a
|
[ROCm][DSV4] Use aiter mHC pre/post as the default ROCm path (#43950)
Signed-off-by: Fangzhou Ai <fangzhou.ai@amd.com>
Signed-off-by: Fangzhou-Ai <fangzhouai@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Claude <noreply@anthropic.com>
|
2026-07-01 16:27:42 +08:00 |
|
|
almayneandLi, Jiang <jiang1.li@intel.com>
|
89e99202f2
|
[CPU][Perf]Added tanh AOR for faster gelu activations. (#44639)
Signed-off-by: Anna Mayne <anna.mayne@arm.com>
Signed-off-by: almayne <anna.mayne@arm.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
|
2026-06-30 23:24:40 -07:00 |
|
|
Gabriel Wu
|
9a08a5118e
|
fix: skip cooperative top-K on SM120 (#47164)
Signed-off-by: Zihua Wu <13583761+lucifer1004@users.noreply.github.com>
|
2026-06-30 21:32:54 -07:00 |
|
|
|
c5200d3565
|
[Attention][DSA] support dcp for FLASHINFER_MLA_SPARSE (#46076)
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
Signed-off-by: Jingyi Yang <girasoleyang@gmail.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: GirasoleY <girasoleyang@gmail.com>
Co-authored-by: Jingyi Yang <girasoleyang@gmail.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
2026-07-01 00:32:20 -04:00 |
|
|
Joe RowellandRobert Shaw
|
c8d2f3cb14
|
[Bugfix] compressed-tensors: allow int8 grouped WNA16 MoE on Marlin (#47154)
Signed-off-by: Joe Rowell <joerowell4@gmail.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
|
2026-06-30 12:50:46 -07:00 |
|
|
Andreas Karatzas
|
c8f9c156a5
|
[ROCm][V1][MLA] Clone prefill backend state per metadata builder (#46993)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
|
2026-06-30 11:43:54 -07:00 |
|
|
|
245888ff77
|
[Feature] Detect all2all peer fault with fault tolerance backend and prevent corrupted output (#43637)
Signed-off-by: fangyuchu <fangyuchu@qq.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
|
2026-06-30 09:00:25 -07:00 |
|
|
Kunshang JiandHarry Mellor
|
e840f0d3f5
|
[Platform] Replace torch.cuda.Event with torch.Event (#47140)
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
|
2026-06-30 08:39:59 -07:00 |
|
|
Arsalan Shakil
|
00ebf19cca
|
[Bugfix][Quant] Raise actionable error instead of bare assert for group-size/TP mismatch (#46230) (#46236)
Signed-off-by: Arsalan Shakil <shakil.arsalan@yahoo.com>
|
2026-06-30 14:57:14 +00:00 |
|
|
Harry Mellor
|
1ab9522935
|
Remove more unnecessary load_weights methods (#47058)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
|
2026-06-30 15:22:16 +01:00 |
|
|
|
8cf7c4d8ad
|
[Attention Backend] add HPC-Ops Attention backend (#46020)
Signed-off-by: chengvjiang <chengvjiang@tencent.com>
Co-authored-by: chengvjiang <chengvjiang@tencent.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>
|
2026-06-30 18:17:43 +08:00 |
|
|
Uros Markovic
|
81bcced482
|
[Bugfix][ROCm] Preserve MoE weight padding for unquantized Triton path (#46381)
Signed-off-by: Uros Markovic <umarkovi@amd.com>
|
2026-06-30 14:47:57 +08:00 |
|
|
Summer YangandClaude
|
5b4cb69523
|
[Bugfix][MLA] Fix LSE log-base mismatch in DCP + FlashInfer MLA decode (#47079)
Signed-off-by: girasoley <girasoleyang@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>
|
2026-06-29 19:15:02 -07:00 |
|
|
Thien Tran
|
43916891b2
|
[GDN] Improve kkt kernel of CuteDSL prefill backend (#46346)
Signed-off-by: Thien Tran <gau.nernst@yahoo.com.sg>
|
2026-06-29 18:34:18 -07:00 |
|
|
Andreas KaratzasandTyler Michael Smith
|
53f7553f09
|
[ROCm][DeepEP] Stabilize high-throughput DBO for DP+EP (#46990)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
|
2026-06-29 14:28:02 -07:00 |
|
|
Jason LiandXin Li
|
debec6440b
|
Add MiniMax-M3 modelopt nvfp4 support (#46756)
Signed-off-by: Xin Li <xinli@nvidia.com>
Signed-off-by: jasonlizhengjian <jasonlizhengjian@gmail.com>
Co-authored-by: Xin Li <xinli@nvidia.com>
|
2026-06-29 09:29:39 -07:00 |
|
|
HDCharles
|
379acd4e4f
|
[Bugfix][Quantization] Fix W8A8 int-quantized scheme selection regression (#46860)
Signed-off-by: HDCharles <charlesdavidhernandez@gmail.com>
|
2026-06-29 15:55:42 +00:00 |
|
|
bnellnmandRobert Shaw
|
bc8481af09
|
[MoE Refactor] Standardize Humming MoE experts + utilities (#43373)
Signed-off-by: Bill Nell <bnell@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
|
2026-06-29 06:19:29 -07:00 |
|
|
Alden LoboandAlden Lobo
|
e186107870
|
[Bugfix] Use native SiLU activation in CPU fused MoE (#45961)
Signed-off-by: Alden Lobo <alden.lobo@arm.com>
Co-authored-by: Alden Lobo <alden.lobo@arm.com>
|
2026-06-29 09:12:20 +00:00 |
|
|
JohnandHarry Mellor
|
0e207dac78
|
[Bugfix] Transformers backend: apply learned lm_head.bias for tied-embedding models (#46835)
Signed-off-by: John Langford <jl@hunch.net>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
|
2026-06-29 08:59:15 +00:00 |
|
|
Harry Mellor
|
5051698e41
|
Remove unnecessary load_weights methods (#44589)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
|
2026-06-29 01:52:23 -07:00 |
|
|
Yuwen Zhou
|
58d6a6e60a
|
[CPU] Support cpu compressed-tensor w8a8 int8 moe (#42920)
Signed-off-by: yuwenzho <yuwen.zhou@intel.com>
Signed-off-by: Yuwen Zhou <yuwen.zhou@intel.com>
|
2026-06-29 03:04:05 +00:00 |
|
|
Ilya MarkovandMarkov Ilya
|
a2abce646f
|
[EPLB] Mask padding in EPLB load recording (#38128)
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Markov Ilya <markovilya19@gmail.com>
Co-authored-by: Markov Ilya <markovilya19@gmail.com>
|
2026-06-28 19:43:58 -07:00 |
|
|
|
4dfbf1503b
|
[Model] Add support for openai/privacy-filter (#41026)
Signed-off-by: Fabian Joswig <fjosw@users.noreply.github.com>
Co-authored-by: wang.yuqi <yuqi.wang@daocloud.io>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
|
2026-06-28 16:18:22 -07:00 |
|
|
fxmarty-amdandAndreas Karatzas
|
03c6d01c30
|
[OCP MX ] Add back emulation to available OCP MX backends list (#46629)
Signed-off-by: Felix Marty <Felix.Marty@amd.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>
|
2026-06-28 12:43:19 -05:00 |
|
|
Woosuk Kwon
|
89876b0c54
|
[GLM5] Implement op fusion for GLM5/DSV3.2 (#46876)
|
2026-06-28 08:17:39 -07:00 |
|
|
|
c6741b2ad4
|
[Model] Support Unlimited OCR (#46564)
Signed-off-by: Tianyu Guo <guoty9@mail2.sysu.edu.cn>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Roger Wang <hey@rogerw.io>
|
2026-06-27 23:09:18 -07:00 |
|
|
jj shaoandshaojinjie.sjj
|
ea2ead1db3
|
[Misc] Fix incorrect layer type annotation in Fp8LinearMethod (#46818)
Signed-off-by: shaojinjie.sjj <shaojinjiesjj@gmail.com>
Co-authored-by: shaojinjie.sjj <shaojinjiesjj@gmail.com>
|
2026-06-27 20:23:59 +00:00 |
|
|
|
867fd5e8ed
|
[ROCm][Perf] Use flydsl moe with Minimax-M3 mxfp8 weights on gfx950 and implemented moe-backend selection (#46184)
Signed-off-by: Hongxia Yang <hongxia.yang@amd.com>
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>
Co-authored-by: Tan Pin Siang <tanpinsiang@gmail.com>
|
2026-06-27 10:22:57 +00:00 |
|
|
Wentao Ye
|
b588f66dc2
|
[GLM5.2 Perf] fused_indexer_q_rope_quant triton kernel, 1.9% ~ 3.3% E2E Throughput improvement. (#46862)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
|
2026-06-26 22:16:20 -07:00 |
|
|
|
ddd3855a28
|
[MoE Backend] add HPC-Ops MoE backend (#45924)
Signed-off-by: chengvjiang <chengvjiang@tencent.com>
Co-authored-by: chengvjiang <chengvjiang@tencent.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>
|
2026-06-27 11:18:07 +08:00 |
|
|
fxmarty-amd
|
37ce34922f
|
[CI] Fix failing CUDA graph capture in Triton MOE (#46735)
Signed-off-by: Felix Marty <Felix.Marty@amd.com>
|
2026-06-26 07:21:20 -07:00 |
|
|
Hongxia YangandClaude Opus 4.8
|
c2507fb293
|
[ROCm] [MoE] [Perf] Shared-expert fusion for bias-routed MoE; enable on MiniMax-M3 mxfp8 model (#46545)
Signed-off-by: Hongxia Yang <hongxia.yang@amd.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
|
2026-06-26 07:05:20 -07:00 |
|
|
TJian
|
8921c4be88
|
[ROCm] [Performance] Optimize aiter moe for DeepSeekV4 (#46122)
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
|
2026-06-26 06:43:27 -07:00 |
|
|
qli88andTJian
|
8e394244a5
|
[ROCm]Enable AITER MoE backend for MiniMax-M3-MXFP4 (#46419)
Signed-off-by: Qiang Li <qiang.li2@amd.com>
Co-authored-by: TJian <tunjian.tan@embeddedllm.com>
|
2026-06-26 06:35:35 -07:00 |
|
|
fxmarty-amd
|
552a9dbe59
|
[NVFP4][Emulation] Fuse NVFP4 weight dequantization with compute in triton kernel for w13/w2 MOE MLP linears (#44667)
Signed-off-by: Felix Marty <Felix.Marty@amd.com>
|
2026-06-25 19:33:00 -07:00 |
|
|
Mike GandMichael Goin
|
ad28d605e6
|
[Bugfix] Default tie_weights to sharing the weight (fix tied quantized embeddings, e.g. ModelOpt Gemma4) (#45544)
Signed-off-by: Mike G <180722391+mikekg@users.noreply.github.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
|
2026-06-25 17:46:28 -07:00 |
|
|
Michael Goin
|
e8e7b592d1
|
[Kernel][MoE] Tune block-FP8 fused MoE for low-batch decode (#46642)
Signed-off-by: mgoin <mgoin64@gmail.com>
|
2026-06-25 12:38:28 -07:00 |
|
|
Qiuyang Yueandmergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
|
9bfd878a48
|
[MoE] [MoE Refactor] Add moe kernel oracle abc 37753 (#43461)
Signed-off-by: Qiuyang Yue <yueqiuyang1389@gmail.com>
Signed-off-by: qyYue1389 <yueqiuyang1389@gmail.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
|
2026-06-25 09:34:03 -04:00 |
|
|
|
710ebaa189
|
[ROCm][Bugfix] Fix chunk alignment when using context parallelism with TRITON_MLA (#46114)
Signed-off-by: Micah Williamson <micah.williamson@amd.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
|
2026-06-25 00:07:28 -04:00 |
|
|
Xin Yang
|
1273a8f05a
|
[Kernel] Add swap AB optimization to fused_moe_kernel (#36559)
Signed-off-by: Xin Yang <xyangx@amazon.com>
|
2026-06-25 01:44:30 +00:00 |
|
|
Kaihang Jiang
|
fc7fc421e9
|
[Kernel][MoE] Allow FlashInfer MXINT4 MoE for gated SiLU (#46518)
Signed-off-by: Kaihang Jiang <kaihangj@nvidia.com>
|
2026-06-24 18:32:50 -05:00 |
|
|
Woosuk KwonandClaude
|
d7ab9be775
|
[Bugfix] Support -1 (invalid/non-local) slots in topk_ids for Triton MoE (#46408)
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
Co-authored-by: Claude <noreply@anthropic.com>
|
2026-06-24 13:59:42 -07:00 |
|
|
Woosuk KwonandClaude
|
6a1570711c
|
[Bugfix] Support non-power-of-2 top_k in legacy triton_kernels routing (#46406)
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
Co-authored-by: Claude <noreply@anthropic.com>
|
2026-06-24 13:52:09 -07:00 |
|
|
|
49f2104c53
|
[Feature] Support DCP with FP8 KV cache in MLA decode path (#44044)
Signed-off-by: shivampr <shivampr.dev@gmail.com>
Signed-off-by: Shivam <shivamprasad91@gmail.com>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
|
2026-06-24 19:28:17 +00:00 |
|
|
HDCharlesandClaude Opus 4.6
|
56ca5997ea
|
Humming support for 2/3/5/6/7-bit pack-quantized weight-only inference (#46389)
Signed-off-by: HDCharles <charlesdavidhernandez@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
|
2026-06-24 13:53:54 -04:00 |
|
|
Fadi Arafeh
|
061043eaca
|
[CPU][Perf] Accelerate unquantized MoE for AArch64 (#46353)
Signed-off-by: Fadi Arafeh <fadi.arafeh@arm.com>
|
2026-06-24 14:14:35 +00:00 |
|